jgchaparro
commited on
Update README.md
Browse files
README.md
CHANGED
@@ -1,7 +1,8 @@
|
|
1 |
---
|
2 |
datasets:
|
3 |
- jgchaparro/language_garden-tsd
|
4 |
-
language:
|
|
|
5 |
model_type: translation
|
6 |
task: translation
|
7 |
---
|
@@ -12,22 +13,22 @@ task: translation
|
|
12 |
|
13 |
|
14 |
|
15 |
-
|
16 |
|
17 |
-
|
|
|
|
|
|
|
18 |
|
19 |
-
|
20 |
|
21 |
-
|
22 |
-
|
23 |
-
This model translates from Greek to Tsakonian. To use it, employ the following prompt:
|
24 |
|
25 |
```plaintext
|
26 |
Translate the following sentence from Greek to Tsakonian: ...
|
27 |
```
|
28 |
|
29 |
-
This model is part of the Language Garden series, a collection of models aimed at preserving endangered languages.
|
30 |
-
|
31 |
## Evaluation
|
32 |
|
33 |
Test set loss: 0.8856971859931946
|
|
|
1 |
---
|
2 |
datasets:
|
3 |
- jgchaparro/language_garden-tsd
|
4 |
+
language:
|
5 |
+
- el
|
6 |
model_type: translation
|
7 |
task: translation
|
8 |
---
|
|
|
13 |
|
14 |
|
15 |
|
16 |
+
# Language Garden: Tyros 8B – Greek to Tsakonian** is an open-source fine-tuned LLM focused on translation from Modern Greek to Tsakonian Greek, the last living descentant of the Doric Greek spoken in Spartan, spoken in the Eastern coast of the Peloponnese.
|
17 |
|
18 |
+
* Developed by Tsakonian Digital.
|
19 |
+
* Trained using the [Meta-Llama-3.1-8B-bnb-4bit](https://huggingface.co/unsloth/Meta-Llama-3.1-8B-bnb-4bit) model and fine-tuned using the [SFT](https://huggingface.co/unsloth/SFT) framework.
|
20 |
+
* Trained on Tsakonian texts from various online sources. Special thanks to Panos Marneris for his efforts on publishing Tsakonian texts.
|
21 |
+
* Training texts use Kostakis' orthography.
|
22 |
|
23 |
+
# Usage
|
24 |
|
25 |
+
Use the following prompt:
|
|
|
|
|
26 |
|
27 |
```plaintext
|
28 |
Translate the following sentence from Greek to Tsakonian: ...
|
29 |
```
|
30 |
|
31 |
+
This model is part of the Language Garden series, a collection of models aimed at preserving endangered languages.
|
|
|
32 |
## Evaluation
|
33 |
|
34 |
Test set loss: 0.8856971859931946
|